A user on the Level1Techs forum reported that Unsloth's default installation failed to run out of the box on an Intel Arc B70 GPU under Fedora 44, and was advised to manually compile the SYCL backend,…
Ternary Bonsai 2 27B, a ternary-weight multimodal model built on a Qwen3.8 27B base, ships at roughly 6GB with weights restricted to {-1, 0, +1} at 1.76 bits per weight with FP16 group scales, accordi…
Prism ML Inc. announced Thursday the launch of Bonsai 2 27B, a ternary-quantized multimodal model built on Qwen3.8 27B that compresses the model from roughly 56 gigabytes at 16-bit to about 5.9 gigaby…
PrismML released Bonsai 2 27B, a ternary-weight compression of Qwen3.8 27B that cuts model size from about 56 GB to 5.9 GB while retaining roughly 98% of the original model's benchmark score. In a han…
PrismML released Ternary Bonsai 2 27B on September 17, an open-weight model built on Alibaba's Qwen3.8 27B that compresses 27.8 billion parameters to roughly 1.76 bits per weight and a 5.9-gigabyte fi…
PrismML released Bonsai 2 27B on Thursday, a compressed version of Alibaba's Qwen3.8 27B open-source model shrunk to 5.9 GB — a 9x to 10x memory reduction that matches 98% of Qwen's aggregate benchmar…
PrismML launched Bonsai 2 27B on September 17th, an Apache 2.0-licensed ternary-weight version of Qwen3.8 27B whose language model fits in 5.95 GB and, according to the company, preserves 98.2% of the…
Edge0, an open-source framework posted to arXiv on September 16, 2026, reports running a 4-bit Qwen3.6-35B-A3B mixture-of-experts model at 20.4 tokens per second inside 2.9 GiB of peak active memory o…
Intel released OpenVINO 2026.4 on 16 September 2026, adding ten new model families for CPU and GPU inference, including Kokoro-82M text-to-speech, Qwen3-ASR speech-to-text, Muse Glimmer 30B, Qwen3.8 2…
AI Model Radar published measurements on 16 September 2026 showing that the standard rule-of-thumb for estimating context-cache memory in open-weights models can be off by a factor of four, and by rou…
A calculator built by Sunk Cost estimates that a $3,499 Apple M5 Max Mac Studio with 64 GB of unified memory running Alibaba's quantized Qwen3.8 27B model would need 44 years to break even against hos…
Two of three new AI trading bots made their scheduled start on the weekly AI scorecard for September 7–13, 2026, with Qwen3.8 27B's @qwen38_daily debuting at #4 with a +37.9% annualized rate on 35 cal…
Abacus.AI released three open-weight Smaug models — Smaug Flash on DeepSeek V4 Flash, Smaug Mini on Qwen3.8 27B, and Smaug Agentic on Kimi K3 — tuned for long-running agent loops and published on Hugg…
A user testing a multi-GPU LLM development machine reported that the system held up well under sustained load running the Qwen3.8 27B model, despite the machine shipping from 45 Drives with the wrong …
Das Lab at the Institute of Science and Technology Austria (ISTA), the team behind GPTQ, introduced GSQ (Gumbel Softmax Quantization) and RCO (Riemannian Constrained Optimization), two linked techniqu…
A new benchmark shows that the Qwen3.8 27B model outperforms GPT 5.6 Luna on the DABstep SQL benchmark while costing under 50 cents in electricity, over 17x less, according to a blog post by Alex Mona…
A hands-on benchmark of Qwen3.8 27B on an RTX 4090 found that 4-bit quantization (NF4 and AWQ INT4) preserves near-baseline quality, with MMLU scores of 59.8% and 59.5% versus 61.2% for FP16, while 1-…
A developer has open-sourced SIMURG, a tool that monitors OpenAI-compatible endpoints and retries requests when it detects hallucinations. Testing with a 2-bit quantized Qwen3.8 27B model on an RTX 30…
A developer's benchmark of Qwen3.8 27B quantizations found that the 17 GB Q4_K_M 4-bit version matches the full BF16 model on Terminal-Bench 2.1, while 1-bit versions collapse to near-random performan…
In a head-to-head test, Anthropic's Claude Opus 5 and Alibaba's Qwen3.8 27B (running locally on an NVIDIA RTX 5090) analyzed the validity of Apple's US Patent 7,844,915 (the 'pinch-to-zoom' patent) ag…